Fuzzy Information Granulation and ED-LSTM based Traffic Prediction of Industrial Control Systems

The industrial control system (ICS) is facing increasing threats in underline communication infrastructure. The mathematical model of communication network traffic in ICS plays a crucial part in the precaution of the cyberattacks. To this end, this paper proposes an integrated prediction approach using learning rate exponential decay (ED-LSTM) method and fuzzy information granulation. The proposed prediction approach is designed to characterize the traffic patterns of ICS for both point and interval prediction. The traffic pattern prediction is essential to characterize the operation behaviors and unique traits in ICS. The experiments and numerical results demonstrate that the proposed integrated prediction approach outperforms the other prediction models in both point prediction and interval prediction.